MeLOn
example_training_of_ANN_with_pruning.py File Reference

Training of artificial neural network in Keras with pruning and export to file that is readable by MeLOn. More...

Namespaces

 example_training_of_ANN_with_pruning
 

Variables

string example_training_of_ANN_with_pruning.problem_name = "peaks"
 LOAD DATA ############################ enter data set information. More...
 
string example_training_of_ANN_with_pruning.filename_data = "./data/peaks.csv"
 
int example_training_of_ANN_with_pruning.input_dim = 2
 
int example_training_of_ANN_with_pruning.output_dim = 1
 
bool example_training_of_ANN_with_pruning.scaleInput = True
 
bool example_training_of_ANN_with_pruning.normalizeOutput = True
 
 example_training_of_ANN_with_pruning.data = np.loadtxt(open(filename_data, "rb"), delimiter=",")
 
 example_training_of_ANN_with_pruning.X = data[:, :-output_dim]
 
 example_training_of_ANN_with_pruning.y = data[:, input_dim:]
 
 example_training_of_ANN_with_pruning.X_norm = utils.scale(X, scaleInput)
 
 example_training_of_ANN_with_pruning.y_norm = utils.normalize(y, normalizeOutput)
 
 example_training_of_ANN_with_pruning.x_train
 
 example_training_of_ANN_with_pruning.x_val
 
 example_training_of_ANN_with_pruning.y_train
 
 example_training_of_ANN_with_pruning.y_val
 
 example_training_of_ANN_with_pruning.test_size
 
 example_training_of_ANN_with_pruning.n_train = x_train.shape[0]
 
string example_training_of_ANN_with_pruning.output_folder = "./data/Output/"
 SET PARAMETERS ############################ output filename. More...
 
string example_training_of_ANN_with_pruning.filename_out = output_folder + problem_name
 
list example_training_of_ANN_with_pruning.network_layout = [10, 10]
 
string example_training_of_ANN_with_pruning.activation_function = 'relu'
 
string example_training_of_ANN_with_pruning.activation_function_out = 'linear'
 
float example_training_of_ANN_with_pruning.learning_rate = 0.001
 
 example_training_of_ANN_with_pruning.kernel_regularizer = tf.keras.regularizers.l2(l=0.0001)
 
string example_training_of_ANN_with_pruning.kernel_initializer = 'he_normal'
 
string example_training_of_ANN_with_pruning.optimizer = 'adam'
 
int example_training_of_ANN_with_pruning.epochs = 100
 
int example_training_of_ANN_with_pruning.batch_size = 128
 
float example_training_of_ANN_with_pruning.initial_sparsity = 0.0
 
float example_training_of_ANN_with_pruning.final_sparsity = 0.4
 
int example_training_of_ANN_with_pruning.begin_step = 30
 
int example_training_of_ANN_with_pruning.end_step = np.ceil(n_train / batch_size).astype(np.int32) * epochs
 
int example_training_of_ANN_with_pruning.frequency = 10
 
int example_training_of_ANN_with_pruning.random_state = 1
 
dictionary example_training_of_ANN_with_pruning.pruning_params
 BUILD MODEL ############################ Set pruning parameters
More...
 
 example_training_of_ANN_with_pruning.pruned_model = tf.keras.Sequential()
 
 example_training_of_ANN_with_pruning.loss
 
 example_training_of_ANN_with_pruning.metrics
 
 example_training_of_ANN_with_pruning.training_time = time.time()
 TRAINING ############################. More...
 
 example_training_of_ANN_with_pruning.history
 
 example_training_of_ANN_with_pruning.stripped_model = sparsity.strip_pruning(pruned_model)
 SAVE MODEL ############################. More...
 
 example_training_of_ANN_with_pruning.y_pred = stripped_model.predict(X_norm)
 

Detailed Description

Training of artificial neural network in Keras with pruning and export to file that is readable by MeLOn.

==============================================================================
Aachener Verfahrenstechnik-Systemverfahrenstechnik, RWTH Aachen University
==============================================================================

Author
Artur M. Schweidtmann, Friedrich von Bülow, Jing Cui, Laurens Lueg, and Alexander Mitsos
Date
20. January 2020